# Fantasy Football Rankings are a Joke. Use Projections Instead!

28Sports Illustrated just released their annual fantasy football rankings. I’ve been waiting in anticipation to find out who they think is the best quarterback in the land. According to SI, it’s Aaron Rodgers. I’ll show why that’s not necessarily correct or even good advice. In this post, I demonstrate the problems with rankings and the solution.

## Problem 1: Rankings are specific to league settings

SI’s rankings are specific to standard league settings. If you use any non-standard scoring settings in your league (e.g., PPR or bonus points for QBs), then their rankings aren’t relevant. Even if you use standard scoring settings, though, rankings are still next to useless because they have another problem.

## Problem 2: Rankings don’t tell you *how much* players are better than each other

Let’s assume we are considering three players: Aaron Rodgers, Peyton Manning, and Drew Brees (ranked as the #1, 2, and 3 QBs, respectively, according to SI). Rankings don’t provide a sufficient basis to know which of these QBs to draft. Who you should draft should be based on the players’ value (in order to compare players across positions) and how much dropoff there is in projected points to the later picks.

For example, let’s assume that Rodgers, Manning, and Brees are projected to score the following points:

- Rodgers: 300 points
- Manning: 298 points
- Brees: 250 points

If we go purely based on rankings, that would mean drafting Rodgers. If we look at projected points, however, Rodgers is only projected to score 2 more points than Manning, who is projected to score 48 more points than Brees. Based on the small dropoff from Rodgers to Manning, and the large dropoff from Manning to Brees, it makes better sense to skip Rodgers (and to draft another position instead) and then, in a later pick, to draft Manning.

## Solution: Use projections

1) Projections can be adapted to your league settings

With projections for passing yards, TDs, INTs, and rushing/reception yards and TDs (and sometimes other categories), we can calculate the projected points for any league. Thus, we can calculate custom projections that are customized for your league.

2) Projections tell you *how much* players are better than each other (dropoff)

By calculating projected points for your league, we can calculate how much dropoff there is from one player to the next. This helps us determine the best players to draft now and who we should draft later.

3) You can calculate rankings from projections

By defining a “typical/baseline replacement player” (see here), we can calculate rankings from projections. For more info on how to do this, see here: http://fantasyfootballanalytics.net/2014/06/custom-rankings-and-projections-for-your-league.html. You can’t go the other way around, however–you can’t calculate projections from rankings.

4) You can account for variability risk and injury risk when using projections

We calculate estimates of source-to-source variability and injury risk in our tools. Another source of variability one could include would be projected game-to-game variability. Performance variability is easier to quantify with projections than rankings.

5) Projections are more accurate than rankings

We have shown that projections are more accurate than rankings.

## Conclusion

Projections are better than rankings because they can be adapted to your specific league settings and they tell you *how much* players are better than each other. Moreover, you can always calculate rankings from projections, but you can’t reverse engineer projections from rankings. The bottom line is that projections are more accurate than rankings. These principles apply to both snake and auction draft leagues. For our app that calculates custom rankings and projections for your league and takes into account dropoff and risk, see here:

Hey Isaac, thanks for putting this together! This tool has been crazy helpful for me so far in mocks, I can’t wait to use it in my live drafts over the next two weeks.

Two quick questions:

1) If I am playing in a 10-team league, what should the Typical Replacement values be set to for each position?

2) Is there any way you could add a “Updated As of (Date)” to your snake draft optimizer? That’d be ideal for understanding where the optimizer is drafting Le’Veon Bell, Josh Gordon etc.

Thanks again man, this is probably the best thing on the internet right now 🙂

Riley

Hey Riley,

1) This website gives an overview of different methods for determining typical replacement (http://www.footballguys.com/05vbdrevisited.htm): average starter, worst starter, players drafted within first 100 picks, etc. They recommend the number of players drafted at each position within the first 100 picks. Another option is the “man games” approach (discussed in this thread: http://fantasyfootballanalytics.net/2013/09/win-your-fantasy-football-snake-draft.html#comment-11470). I don’t know which this the best to use, but we have plans to compare their accuracy next year.

2) This is a great suggestion. I don’t see an easy implementation, however, because the scripts don’t tell us when each site last updated their data. So far as I can tell, this would have to be entered manually (i.e., going to each site and writing it down). I’ll look more into this because I like this idea, but I probably won’t get to it this season. In the meantime, I’d recommend you go to each site to see when they last updated their projections if you want to alter weights accordingly.

-Isaac

Thanks! Great article on VBD; I dug around for more info on calculating Baseline numbers for VBD drafting and found an article from Chris Harris at ESPN saying to use “Top 100” method for 10-team leagues, and “Top 120” for 12-team leagues…interesting stuff.

Anyways, love what you’re doing here, I work in finance and I explained your website to a friend of mine the other day as factor based passive investing with a value tilt. He understood immediately.

I have a theory on another factor to consider: if it comes down to two players with virtually identical VOR, roughly equal risk, and you don’t particularly care for one position over the other, I think I prefer the player on a team with higher overall points projected by the masses. i.e. players on Denver, New Orleans, Green Bay, New England have a slight edge since they’re on a team with higher projected points. However, maybe this is just double counting some of their value?

Thanks for the Top 120 method suggested by Chris Harris for 12-team leagues. Just added it as a possibility for determining “typical replacement” in the calculation of VOR (http://fantasyfootballanalytics.net/2013/04/win-your-snake-draft-calculating-value.html). Fantasy football has a lot of similarities to finance. I note some of the similarities in this post: http://fantasyfootballanalytics.net/2014/06/best-fantasy-football-projections-2014.html.

Interesting possibility that a player’s offense could be a tie-breaker when selecting between players with similar VOR. Hopefully the strength of a player’s offense would be accounted for in the original player’s projection. If so, it would be double counting the offense strength to consider it on top of the projection. If the projections don’t consider the surrounding offense, however, then this would be unique and important information. I suspect that websites’ projections take offense into consideration somewhat, but they’re mostly black boxes… we don’t know what inputs they use. Maybe one of these days we can generate our own projections from scratch using a transparent, community-driven approach.

So based on your article the most accurate 5 dollar ap i could buy would be rotowires projection based cheat sheet?

Hi John,

The most accurate app would be this (free) tool that uses the wisdom of the crowd for calculating projections that are customized to your league settings:

http://apps.fantasyfootballanalytics.net/projections

-Isaac

Isaac, I am so glad I stumbled on to this via Google search. This is an incredible tool. I love the dropoff assesment column and just the fact I can customize the stats to my league. You should try and sell this to Yahoo, CBS or any of the others that don’t have the customize option.

I am in a two QB league. Typically the QB’s will go in the top 6 and then the top RB’s and WR and then more QB’s. I’m not sure but if I double all of the QB’s value would that make it a good comparison? Probably not because some of the mid tier QB’s are not more valuable than the better RB’s WR or J Graham.

Thanks for the awesome info.

Hey Rob,

Unless your scoring settings heavily favor QBs, you’re probably best drafting QBs after RBs, WRs, and TEs. There’s good value to be had in QBs later in in the draft (http://fantasyfootballanalytics.net/2013/07/expected-points-by-position-rank-in-fantasy-football.html). I wouldn’t artificially inflate QBs’ value–the scoring settings should determine value. If the VOR statistics suggest drafting other positions, then you have identified a bias in your competitors’ drafting styles that you can exploit!

-Isaac

Hey Isaac. Realize this is a year late, but wanted to respond to your article. See, I’ve always thought that rankings are far more valuable than projection and that projections are, for lack of a better term, bullshit. One of the biggest problems with projection comes with suspended players, like Tom Brady. Tom’s projections likely place him much lower than his actual value, because the projections don’t factor in the replacement-level quarterback points you will get in your lineup with Brady suspended. It is also very difficult for projections to factor in injury risk, because most projections assume players play the full 16 game season. And, if they do factor in injury risk, and project a player for 13 game season, projections run into the above suspension problem. Finally, projections have difficulty encompassing the range of a player’s potential. For example, CJ Anderson could have 1400 yards and 12 tds, or he could lose his job and get 800 yards. An average of those two scenarios is a poor reflection of CJ’s value. I think the best method is rankings accompanied by auction values to show the difference between different ranked players.

Hi Eric,

Projections can take into account how many games a player is projected to play (based on suspensions, injuries, etc.), so projections absolutely can factor in replacement-level points. We have plans to include this. We already account for injury risk and players’ ceiling/floor. Ultimately, whether rankings or projections are more accurate is a testable question that we plan to test.

Thanks,

Isaac

Thanks Isaac for putting this together. This brings together two of my passions – fantasy football and data analytics.

I wanted to know if there was a way to get a recommended draft pick order by position relative to location in the draft order for my snake draft on a standard scoring espn league.

Essentially which position to pick first (RB, WR, QB, TE) based on what my position on the draft order is. If I am #1 would I pick the highest ranked RB or the highest ranked WR and so on till my roster and bench is full.

In order to answer this – I was thinking that if I can look at the pattern of fantasy football winners across ESPN or other sources for the past 3 years within leagues of similar format and size (>10 team leagues minimum) and look at the draft order they took based on where their drafting position in the snake draft was – I could use the same pattern for my own snake draft and focus on filtering for the right position/player available.

So for instance if I went 4 in round one and 8 in round two and based on the the larget %-age of population above in the same slots the recommendation was RB (first pick) and QB (second pick), then I could look for best available RB when my turn came and best QB the next go around.

I would ideally like to go into the draft engine and set Autopick with Best player per ADP or projections by position – so pick one I would filter for projected metricand pick the best RB available and for pick two filter for QB and pick the best QB available and so on.

I think in Fantasy Football – access to the position order for snake draft is a higher value as a ‘cheat sheet’ than individual rankings or depth charts (which are useful for sure). Plus with the access to data on all leagues in the system that order would refine over time and be reflective of the league at large and incorporate crowd knowledge into the results.

Please let me know if you already have something that I can churn out a cheat sheet with. Because that would be AWESOMMMMEEE!!!

Something like this (Position n represents my potential position in the snake draft relative to everyone else in my league):

Position 1 – RB1, QB1, WR1, RB2, QB2, WR2, TE1….

…

…

Position 12 – RB1, QB1, WR1, RB2, QB2, WR2, TE1…

Not that it should impact you at all but thought I’d mention that my draft is on August 19 at 12:15pm

Details: ESPN Standard scoring PPR league

Position Limits:

QB – 3

RB – 5

WR – 5

TE – 3

Team has 1 QB, 2 RB, 2 WR, 1 Flex, 1 K, 1 D/ST and 5 slots on bench.

🙂

Thanks,

Sandeep

Hi Sandeep,

Check out our article on the Snake tool. See especially the Strategy section to help you determine who to draft and what factors to consider:

http://fantasyfootballanalytics.net/2013/09/win-your-fantasy-football-snake-draft.html

Hope that helps,

Isaac

HI Isaac. I am a huge sports fan and I love fantasy football. I love the way you are using analytics to do fantasy projections. It seems that analytics are becoming more used in every sport. However I am not good at math and I have no clue about analytics. I am in a 16 team ppr league and we use a snake draft. The problem is we have a pretty unique scoring system. I was wondering if I showed you how we score our league, could you please show me how to use the projections app and the snake tool app. I would love to use all the great tools you have on this site. You are obviously light years smarter then me. I would appreciate if I showed you our scoring system could you please tell me if I can use your web site and apps for my league. Thank you so much for creating this site and for all your help. Below I have listed some of the scoring values for my league. If you would be so kind would you take a look at it and let me know if I can use your apps and how to do so. thank you so much for your time. Please email me back I would greatly appreciate it. Thank you.

Offensive Name Settings

FG Field Goals

3 points

Plus 2 points for a FG of 40 to 49 Yds

Plus 4 points for a FG of 50 to 59 Yds

Plus 7 points for a FG of 60 to 99 Yds

FL

Fumble Lost, Including ST plays

-2 points

IFRTD Individual Fumble Recovery TD

6 points

Plus a 3 point bonus @ 10 to 39 IFRTD

Plus a 6 point bonus @ 40+ IFRTD

IKRTD

Individual Kick Return TD

6 points

IPRTD

Individual Punt Return TD

6 points

MFG

Missed Field Goal

-1 point

MXP

Missed Extra Point

-3 points

Pa2P

Passing Two-point Conversion

2 points

PaInt

Passing Interception

-2 points

PaTD Passing TD

6 points

Plus 3 points for a PaTD of 10 to 39 Yds

Plus 6 points for a PaTD of 40+ Yds

PaYd Passing Yards

Plus a 6 point bonus @ 250+ PaYd

Plus a 6 point bonus @ 300+ PaYd

Plus a 6 point bonus @ 350+ PaYd

Plus a 9 point bonus @ 400+ PaYd

Plus a 12 point bonus @ 450+ PaYd

Plus a 15 point bonus @ 500+ PaYd

Re2P

Receiving Two-point Conversion

2 points

ReTD QB

Receiving TD

6 points

Plus 3 points for a ReTD of 10 to 39 Yds

Plus 6 points for a ReTD of 40+ Yds

Recpt

Reception

1 point

Ru2P

Rushing Two-point Conversion

2 points

RuReYd Rushing and Receiving Yards count together QB

Plus a 3 point bonus @ 50+ RuReYd

Plus a 3 point bonus @ 75+ RuReYd

Plus a 6 point bonus @ 100+ RuReYd

Plus a 3 point bonus @ 125+ RuReYd

Plus a 4 point bonus @ 150+ RuReYd

Plus a 5 point bonus @ 175+ RuReYd

Plus a 6 point bonus @ 200+ RuReYd

Plus a 7 point bonus @ 225+ RuReYd

Plus a 8 point bonus @ 250+ RuReYd

Plus a 9 point bonus @ 275+ RuReYd

Plus a 10 point bonus @ 300+ RuReYd

RuTD Rushing TD QB

6 points

Plus 3 points for a RuTD of 10 to 39 Yds

Plus 6 points for a RuTD of 40+ Yds

XP

Extra Points 1 point

RuReYd Rushing and Receiving Yards WR

Plus a 3 point bonus @ 25+ RuReYd

Plus a 6 point bonus @ 50+ RuReYd

Plus a 3 point bonus @ 75+ RuReYd

Plus a 3 point bonus @ 100+ RuReYd

Plus a 4 point bonus @ 125+ RuReYd

Plus a 5 point bonus @ 150+ RuReYd

Plus a 6 point bonus @ 175+ RuReYd

Plus a 7 point bonus @ 200+ RuReYd

RuReYd Rushing and Receiving Yards

Are counted as one group for RB’s

Plus a 6 point bonus @ 50+ RuReYd

Plus a 3 point bonus @ 75+ RuReYd

Plus a 6 point bonus @ 100+ RuReYd

Plus a 3 point bonus @ 125+ RuReYd

Plus a 4 point bonus @ 150+ RuReYd

Plus a 5 point bonus @ 175+ RuReYd

Plus a 6 point bonus @ 200+ RuReYd

Plus a 7 point bonus @ 225+ RuReYd

Plus a 8 point bonus @ 250+ RuReYd

Plus a 9 point bonus @ 275+ RuReYd

Plus a 10 point bonus @ 300+ RuReYd

Defensive

Name

Settings

BFTD Blocked Field Goal Touchdown (ID/ST/DST)

6 points

Plus 3 points for a BFTD of 30 to 49 Yds

Plus 6 points for a BFTD of 50 to 99 Yds

BPTD

Blocked Punt Touchdown (ID/ST/DST)

6 points

Plus 3 points for a BPTD of 30 to 49 Yds

Plus 6 points for a BPTD of 50 to 99 Yds

DFR

Defensive/ST Fumble Recovered (ID/DT/DST)

2 points

DFRTD Defensive Fumble Recovery TD (ID/DT/DST)

6 points

Plus 3 points for a DFRTD of 20 to 39 Yds

Plus 6 points for a DFRTD of 40 to 99 Yds

Int

Interceptions

2 points

IntTD Interception TD

6 points

Plus 3 points for a IntTD of 20 to 39 Yds

Plus 6 points for a IntTD of 40 to 99 Yds

KRTD Kick Return TD (ID/ST/DST)

6 points

Plus 3 points for a KRTD of 60 to 79 Yds

Plus 6 points for a KRTD of 80+ Yds

PA Points Against, Total Points Scored

0 – 0 PAs = 15 points

2 – 5 PA = 10 points

6 – 9 PA = 5 points

10 – 13 PA = 3 points

14 – 34 PA = 0 points

35 – 39 PA = -1 point

40 – 44 PA = -3 points

45 – 49 PA = -5 points

50 – 99 PA = -7 points

PRTD Punt Return TD (ID/ST/DST)

6 points

Plus 3 points for a PRTD of 50 to 69 Yds

Plus 6 points for a PRTD of 70+ Yds

SACK

Sack

1 point

SFRTD

Special Team Fumble Recovery for TD

6 points

STY

Safety

5 points

YDS Yards Allowed

0 – 99 YDSs = 5 points

100 – 199 YDSs = 3 points

200 – 299 YDSs = 1 point

300 – 399 YDSs = 0 points

400 – 499 YDSs = -1 point

500 – 599 YDSs = -3 points

600 – 699 YDSs = -5 points

If you do not understand the scoring system please let me know by email. My brother is the commissioner of the league. If you do not understand the scoring I will get with him and clarify it and email it to you if needed. Thank you very much.

Hi Mike,

You should be able to plug many of those league settings into our app. Some of those scoring settings are fairly specific and not projected by sites, so we couldn’t include them in our projections. You should be able to get a pretty good estimate of players’ projections by plugging in most of those scoring settings. Let me know if you have a question about what a particular scoring setting means—they should be fairly self explanatory, but you can get more info on the Projections and Snake tools here:

http://fantasyfootballanalytics.net/2014/06/custom-rankings-and-projections-for-your-league.html

http://fantasyfootballanalytics.net/2013/09/win-your-fantasy-football-snake-draft.html

Hope that helps,

Isaac

Thank you for replying to my post. If I have any questions or don’t understand anything I’ll let you know. Thank you for your time.

What do you recommend when rankings and projections are telling me different things when comparing two players?

For example, Demaryius Thomas has better projections (higher average, equal ceiling, higher floor) than Dez Bryant, but Dez is ranked two spots higher on FantasyPro’s ECR for WRs. Based on ECR, ADP, and AAV, Dez is expected to be better, but Demaryius has the better projections.

Use Projections. ADP and AAV tells you what the marked on average thinks the players are worth. Furthermore, to quote the above article:

“Projections are better than rankings because they can be adapted to your specific league settings and they tell you how much players are better than each other. Moreover, you can always calculate rankings from projections, but you can’t reverse engineer projections from rankings. ”

If Demaryius is projected more, it means the experts think he will score more points in your league. Thus he is the better player for you. ECR is far less important for the reasons given in the above quote.

-Eirik

But using the rankings for standard scoring, the experts think that Dez is better. Rankings seem to have the advantage that there are significantly more experts doing rankings than projections. The projections app uses projections from 12 sources, while the Fantasypros.com ECR uses data from 49 sources (and up to 121 if you want). Projections are definitely more useful for comparing players across different positions and determining their VOR or creating custom rankings using different scoring rules. But, since I’m just using standard scoring, it seems that the ECR may be better for deciding between two players at the same position.

Is there any data about the accuracy of projections vs rankings when comparing players at the same position?

Hey Issac I don’t mean to be a troll but I’ve been going through your site and I’ve seen some of your projections have been way off. I understand that on average you have the most accurate sight based on your R calculations and that some missed projections are natural. Do you by chance calculate how far off your projections are when you do miss big? I love the website but missing by 3 points is a huge difference from missing by 10+.

Also I’m trying to think how this would be calculated but creating a forecast on the percentage owned a player is for weekly daily fantasy leagues. I’m not sure how to create an accurate prediction of player ownership but I think it could possibly be done by taking %Player-Owned by Teams based on projected favorability of the matchup. I don’t know if you have access to this information but it would be nice to how player rank to opposing defense influences ownership from week to week.

This is a problem of predicting weekly performance, which involves lots of random variability (chance). We aren’t generating our own projections—we are averaging many sources of projections, so our projections will only be as good as (the average of) the sources we include. Nevertheless, we’ve shown that the average is likely more accurate than any individual source of projections. In any case, seasonal projections are likely more accurate than weekly projections. For more info, see here:

http://fantasyfootballanalytics.net/2015/02/stats-help-fantasy-football.html

It’s an interesting idea to see if % ownership would improve performance forecasts above and beyond projections.

What is the r script like is it a program in last and want to get familiar with it

http://fantasyfootballanalytics.net/2014/06/learn-r.html

hi Isaac – i posted the a question several days ago on the appropriate page about a link you posted but didn’t get a response, so i’m reposting here. The following link doesn’t work. Message says “cannot connect to the server”. please advise. thanks.

http://fantasyfootballanalytics.net:3838/Stocks/

We lost Shiny server when we upgraded our servers. It’s on our to-do list.

I couldn’t refrain from commenting. Well written!

Hello Isaac, when do you plan on having 2016 Week 1 Projections available to download from your site?

Thanks!

We should have Week 1 projections up when more sources provide them. Hopefully in the week or two before the season starts!

-Isaac